Computational identification of deleterious synonymous variants in human genomes using a feature-based approach

Area of Science:

• Genomics
• Computational Biology
• Human Genetics

Background:

• Synonymous single nucleotide variants (sSNVs) can impact human disease despite not altering protein sequences.
• Distinguishing pathogenic sSNVs from neutral variants is challenging due to their low prevalence.
• Existing methods for predicting variant impact are limited in their focus on identifying pathogenic sSNVs.

Purpose of the Study:

• To develop a computational model for identifying deleterious synonymous single nucleotide variants (sSNVs).
• To systematically investigate and select informative features for predicting pathogenic sSNVs.
• To compare the performance of the developed model against state-of-the-art methods.

Main Methods:

• A random forest (RF) classifier was developed, named IDSV (Identification of Deleterious Synonymous Variants).
• Seventy-four features across seven categories (splicing, conservation, codon usage, sequence, pre-mRNA folding energy, translation efficiency, function regions annotation) were investigated.
• Feature selection using sequential backward selection identified an optimized subset of 10 features for the RF model.

Main Results:

• The IDSV model, utilizing an optimized feature set, demonstrated superior performance in identifying pathogenic sSNVs on benchmark datasets.
• The study identified translation efficiency as an informative feature for detecting deleterious sSNVs, alongside splicing and conservation features.
• Function region annotation and sequence features showed weaker individual informativeness but potential utility when combined with other features.

Conclusions:

• An efficient, feature-based prediction approach (IDSV) was developed for identifying deleterious sSNVs.
• A compact subset of 10 features, including translation efficiency, splicing, and conservation, is highly effective for sSNV pathogenicity prediction.
• The study provides valuable data and source code for the research community to advance the identification of disease-associated sSNVs.